The episode of the AI Hardware Show explores six leading semiconductor chips designed specifically for the automotive industry, highlighting the unique challenges and innovations in this sector. Sally Wardton and Ian Kudras introduce the discussion by emphasizing the complexity and high demands of AI compute in automotive applications. The episode is sponsored by Arterus, a company that provides automation tools to streamline the design of complex system-on-chips (SoCs) used in everything from data centers to self-driving cars, underscoring the importance of efficient chip design in automotive technology.
Mobileye’s IQ6 chip stands out as a major player, with over 170 million vehicles already equipped with its technology. Built on a 7nm process, the IQ6 supports deep learning for mainstream automotive systems, including advanced driver-assistance systems (ADAS) and hands-free driving. Mobileye’s approach integrates multiple functions into a single SoC, reducing cost and weight while accelerating development. Their roadmap focuses on software-first autonomy with over-the-air updates and scalable features, positioning Mobileye as a key supplier with a strong hardware-software validation track record.
Ambarella’s CB3 family represents a shift from traditional vision chips to full-stack autonomous driving solutions, designed from the ground up for real-time AI workloads like sensor fusion and trajectory planning. Fabricated on a 5nm process, these chips emphasize energy efficiency and low latency, supporting a wide range of vehicle functions beyond driving, such as driver monitoring and radar processing. Despite industry headwinds, Ambarella expects growth driven by the rise of software-defined vehicles, leveraging its specialized hardware architecture tailored to automotive AI needs.
Halo, a startup focused on vision and AI chips, has gained traction particularly in the Chinese EV market with its power-efficient Halo 8 and the newer 10H chip, which supports transformer models and smaller language models for cockpit systems. Halo’s strong software ecosystem and developer community, along with significant funding, enable rapid innovation and deployment. Their chips target a niche for low-power AI inference, highlighting the growing importance of AI in vehicle interiors and user interfaces.
Nvidia’s Thor chip aims to consolidate multiple vehicle functions into a single centralized platform, powered by the new Blackwell GPU architecture and advanced CPU cores. While offering impressive compute capabilities and integration, Thor faces challenges with production delays and the complexity of ensuring safety-critical domain isolation. Meanwhile, NXP focuses on long lifecycle reliability with its IMX95 chip, balancing AI performance with safety and availability for embedded automotive applications. Lastly, Athos Silicon, spun out from Mercedes-Benz, targets level 4 autonomy with a safety-first modular chiplet design using high-bandwidth memory (HBM) for failover and reliability, addressing the stringent demands of automotive safety and longevity. The episode concludes by highlighting the critical role of specialized hardware and software integration in advancing safe, scalable AI for the automotive industry.
